Big Tech Investments Boost Cloud Revenue, Raise AI Risks

Corporate filings show OpenAI and Anthropic account for more than half of roughly $2 trillion in future cloud commitments across Microsoft, Amazon, Google and Oracle. Much of the funding took the form of cloud credits that the startups then used on the same providers.

Providers record that cloud consumption as commercial revenue, while investors can increase the reported value of their equity stakes and record valuation markups on their income statements. Some observers describe the sequence as a round-trip funding loop because the money flows back to the investor as cloud spending and accounting gains. The arrangement is permitted under current accounting rules.

Microsoft’s $13 billion stake in OpenAI was delivered largely as Azure credits. Filings and company disclosures indicate OpenAI’s annual cloud bill has climbed past $60 billion while reported revenue is roughly $25 billion. Anthropic spent about $2.66 billion on Amazon Web Services in nine months, a pace that roughly matched its revenue over the same period.

The accounting effects are visible in recent quarterly results. Alphabet reported $62.6 billion in profit in the first quarter of 2026, with about $28.7 billion attributed to a valuation markup on its Anthropic stake. Amazon’s net income of about $30.3 billion included approximately $16.8 billion tied to Anthropic-related accounting; the company’s free cash flow fell about 95% to $1.2 billion as capital spending on data centers increased.

Microsoft reported that roughly 49% of its $627 billion future backlog is linked to OpenAI, and Oracle shows about 54% of a $553 billion pipeline tied to a single large AI customer.

Companies outside the investor-startup funding loop report rising bills. One ride-hailing company said it exhausted its 2026 AI coding budget by April after deploying Anthropic tools to thousands of engineers, with some staff running monthly API charges in the hundreds or thousands of dollars. Microsoft limited internal use of a third-party coding model after token consumption became expensive for internal budgets.

Bryan Catanzaro, Nvidia’s vice president of applied deep learning, said, “For my team, the cost of compute is far beyond the costs of the employees.” Analysts note that cheaper chips can lower per-unit compute costs but may also encourage heavier usage, which can keep overall spending high.

Market participants describe a “prove-it” phase in which investors and customers are focusing on whether large-scale AI deployments can be sustained under normal commercial terms.